Addressing the challenge of calculating critical rainfall for flash floods in areas without data,a study was con-ducted on 45 small watersheds in mountainous and hilly regions along rivers in Wulian County,Shandong Province,to develop a machine learning-based model for calculating critical rainfall for flash floods.The disaster-causing water levels and critical discharges for each village were determined through a combination of hydrological and hydraulic methods along with on-site investigations.Flood processes were then derived through runoff production and concentration calculations using a hydrological model.The critical rainfall was calculated using a trial and error method.Relevant characteristic para-meters such as hydrological data,underlying surface conditions,and villages along the rivers,along with the calculated critical rainfall results,were selected as training parameters.Based on the extreme gradient boosting algorithm,different prediction models for critical rainfall during different warning periods were constructed.The accuracy of these models was evaluated using the mean absolute error and determination coefficient.The results show that the average absolute errors of the critical rainfall calculated by this model for each warning period are 4.56,6.68,and 7.11,respectively,while the determination coefficients are 0.955,0.967,and 0.973,respectively.The model exhibits high prediction accuracy,meeting the application requirements for flash flood warning in areas without data.
mountain flood disastercritical rainfallextreme gradient boosting algrithmwarning indexWulian County
NETL
NSTL
万方数据
